TIL: Quick & Convenient Cypher Counts

I’m a little late to the party on this one, but since the around the 5.6 release of Neo4j, you are now able to use Cypher Subqueries Neo4j 5.0 to get node and relationship counts from the count store in Neo4j.

For as long as I can remember, I have been writing the following Cypher statement to get a count of nodes with a label.

MATCH (p:Person) RETURN count(p) AS count

As long as you specify a simple one-degree pattern, for example (:Label), ()-[:TYPE]->(), (:Label)--(:Label) or (:Label)-[:TYPE]->(), Neo4j will use the database statistics held in the count store to produce an instant result.

This comes with a caveat, you can only do this once per query. So if you wanted more than one count, you would need to combine counts together using a UNION query.

Fast-forward to 2023

Today I learned that you can use COUNT subquery expressions to do the same thing with less code. Subquery expressions are subqueries that can be added to a WITH or WHERE clause.

So the above query to get a count of all (:Person) nodes now becomes:

RETURN COUNT { (p:Person) }

That’s 42 characters reduced to 28 - 66.666% less. I estimate I write a query like this 5-10 times a day, so over 48 working weeks a year, that means a saving of 67 200 characters a year.

Here is an example that counts the number of :ACTED_IN relationships from a (:Person) node and the total count of :REVIEWED relationships.

RETURN COUNT { (p:Person)-[:ACTED_IN]->() } AS roles,
    COUNT { ()-[:REVIEWED]->() } AS reviews

No More UNIONs

This also means no more unions as each subquery is planned seperately. The following query calculates the percentage of (:HelpfulResponse)s compared to the overall number of (:Response)s.

WITH COUNT { (r:Response) } as total,
     COUNT { (r:HelpfulResponse) } AS helpful
RETURN round(100.0 * helpful / total, 2) +'%'

Thank you to the Cypher team and the #cypher-genuis-bar!